Impact of an integrated nutrition and health programme on neonatal mortality in rural northern India Abdullah H Baqui,a Emma K Williams,a Amanda M Rosecrans,a Praween K Agrawal,a Saifuddin Ahmed,b Gary L Darmstadt,a Vishwajeet Kumar,a Usha Kiran,c Dharmendra Panwar,c Ramesh C Ahuja,d Vinod K Srivastava,d Robert E Black a & Manthuram Santosham a

Objective To assess the impact of the newborn health component of a large-scale community-based integrated nutrition and health programme. Methods Using a quasi-experimental design, we evaluated a programme facilitated by a nongovernmental organization that was implemented by the Indian government within existing infrastructure in two rural districts of Uttar Pradesh, northern India. Mothers who had given birth in the 2 years preceding the surveys were interviewed during the baseline (n = 14 952) and endline (n = 13 826) surveys. The primary outcome measure was reduction of neonatal mortality. Findings In the intervention district, the frequency of home visits by community-based workers increased during both antenatal (from 16% to 56%) and postnatal (from 3% to 39%) periods, as did frequency of maternal and newborn care practices. In the comparison district, no improvement in home visits was observed and the only notable behaviour change was that women had saved money for emergency medical treatment. Neonatal mortality rates remained unchanged in both districts when only an antenatal visit was received. However, neonates who received a postnatal home visit within 28 days of birth had 34% lower neonatal mortality (35.7 deaths per 1000 live births, 95% confidence interval, CI: 29.2–42.1) than those who received no postnatal visit (53.8 deaths per 1000 live births, 95% CI: 48.9–58.8), after adjusting for sociodemographic variables. Three-quarters of the mortality reduction was seen in those who were visited within the first 3 days after birth. The effect on mortality remained statistically significant when excluding babies who died on the day of birth. Conclusion The limited programme coverage did not enable an effect on neonatal mortality to be observed at the population level. A reduction in neonatal mortality rates in those receiving postnatal home visits shows potential for the programme to have an effect on neonatal deaths.

Bulletin of the World Health Organization 2008;86:796–804.

الرتجمة العربية لهذه الخالصة يف نهاية النص الكامل لهذه املقالة. .Une traduction en français de ce résumé figure à la fin de l’article. Al final del artículo se facilita una traducción al español

Introduction infections (36%), complications of ied. However, population-level data on preterm birth (28%) and birth asphyxia the impact of family and community- The neonatal period is recognized as a (23%), with low birth weight as the based neonatal care from large-scale brief, critical time that requires focused primary contributory cause.1 High cov- programmes is scarce. A need exists interventions to reach the Millennium erage of a few simple and cost-effective to implement proven interventions at Development Goal of a two-thirds interventions would reduce neonatal scale, across the continuum of care, reduction in child mortality by 2015. mortality.4–7 Interventions at the fam- without losing impact. In India there are one million neonatal ily and community level can save Here, we evaluate the effect of a deaths every year, representing approxi- lives, especially where health systems community-based package of maternal mately a quarter of all global neonatal are weak.4,8 Several community-based and newborn interventions that was deaths.1,2 Neonatal deaths account for efficacy trials have evaluated service implemented at scale using existing about 38% of the annual 10.6 million delivery strategies to improve newborn government infrastructure through an child deaths recorded worldwide and survival.7,9–16 Those trials were con- integrated nutrition and health pro- nearly half of the deaths in children ducted under controlled conditions to gramme in eight states of India. This under 5 years in India.1,3 ensure high programme coverage and evaluation was conducted in two rural In developing countries, the pri- usually employed workers who deliv- districts of the state of Uttar Pradesh, mary causes of neonatal deaths are ered only the intervention being stud- India.

a Department of International Health, Johns Hopkins Bloomberg School of Public Health, 615 N Wolfe Street, Baltimore, MD 21205, United States of America (USA). b Department of Population and Family Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA. c CARE-India, New Delhi, India. d King George Medical University, Lucknow, India. Correspondence to Abdullah H Baqui (e-mail: [email protected]). doi:10.2471/BLT.07.042226 (Submitted: 21 March 2007 – Revised version received: 12 December 2007 – Accepted: 6 March 2008 )

796 Bulletin of the World Health Organization | October 2008, 86 (10) Research Abdullah H Baqui et al. Nutrition and health programme’s impact on neonatal mortality

Fig. 1. Conceptual model for promotion of newborn care within the INHP

Ministry of Women and Ministry of Health CARE-India Child Development’s Integrated Planning, coordination and logistical and Family Welfare Child Development Services Implementing Infrastructure support to integrate and strengthen partners ICDS and MOHFW programmes with Infrastructure One auxiliary nurse-midwife an emphasis on newborn care and per 5000 One anganwadi worker per Technical support, training of government 1000 (or 1 village) their inputs Planning, training, supervision, officials, frontline health workers Planning, training, supervision, supplies, logistics and community volunteers and monitoring supplies, logistics Funding to NGOs for behaviour change and monitoring Programme and policy communication and social marketing development Programme and policy Monitoring and evaluation development

Health workers’ knowledge improved Increased programme coverage Processes Improved monitoring systems at the block and district level Volunteer change agents recruited and trained; volunteer women’s groups formed at the village level Improved supervision Strengthened supply chain

Mothers’ knowledge Improved use Outcomes Improved preventive care Improved newborn of maternal and newborn practices for mothers care practices of health-care care increased services

Impact Reduced neonatal mortality rate

anganwadi, government child-care and mother-care centre; ICDS, Integrated and Development Services; INHP, Integrated Nutrition and Health Programme; MOHFW, Ministry of Health and Family Welfare; NGO, nongovernmental organization.

Methods bearing women and children under the mentary food to poor families, and age of 5 years; these interventions were provide preschool education.17–19 The Programme description evaluated separately. INHP encouraged anganwadi workers The Integrated Nutrition and Health In both the INHP and standard to recruit community volunteers called Programme (INHP) was a partnership government health services, health “change agents” to further improve the of an international nongovernmental education and services are provided by reach of programmes. The anganwadi organization (NGO), CARE-India, two groups of government functionaries: workers, auxiliary nurse-midwives, and with the Indian government and local auxiliary nurse-midwives, and maternal change agents in the intervention NGOs. The programme was imple- and child health promotion (anganwadi) district received a total of 6 days of mented through the infrastructure of workers. Auxiliary nurse-midwives work training on the care of mothers and the government’s Ministry of Women in health centres that serve a rural popu- newborn babies; as a group, we refer to and Child Development’s Integrated lation of about 5000 people. They make them as community-based workers. In Child Development Services and the home visits to promote home care and the INHP, information to encourage Ministry of Health and Family Welfare care-seeking, attend deliveries, provide behaviour change was usually commu- (Fig. 1). The newborn care package immunization and encourage use of nicated during antenatal and postnatal aimed to increase the frequency of be- family planning methods. Anganwadi home visits by the community-based haviours during the antenatal, delivery workers serve one village (a population workers (Table 1). and postnatal periods that have proven of approximately 1000) and operate benefits for maternal and newborn sur- a facility called an anganwadi centre. Study location, population and vival (Table 1). The INHP also included They promote maternal, newborn design interventions to improve immunization and child health from fixed sites and Although the programme was imple- rates and the nutritional status of child- through home visits, distribute supple- mented in eight states, we collected

Bulletin of the World Health Organization | October 2008, 86 (10) 797 Research Nutrition and health programme’s impact on neonatal mortality Abdullah H Baqui et al.

Table 1. Intervention components of the neonatal care package

Period Strategies related to newborn health Behaviours promoted to mothers/families

Prenatal Home visitation by auxiliary nurse-midwife, aganwadi • Early registration of pregnancy with aganwadi worker and auxiliary worker and change agent to provide counselling on nurse-midwife preventive care, nutrition, preparedness for child • At least three antenatal check-ups birth, and health-care utilization for complications. • Two doses of tetanus toxoid immunization per pregnancy Food supplements provided to poor families at • Daily iron-folic acid supplements for 3 months anganwadi centre. • Reduction of pregnant women’s workload (rest at least 2 hours/day) • Consumption of an additional meal or snack per day and micronutrient rich foods • Birth preparedness: identification of trained provider and a clean delivery site, savings for emergency, arrangement for transport if needed • Obtaining disposable delivery kit or prepare delivery kit • Identify and seek care for danger signs in mothers and neonates Delivery Encourage families to call auxiliary nurse-midwife or • Practice five cleans: clean surface for delivery, clean hands, new blade, trained traditional birth attendant to attend delivery. clean cord tie, clean cloth to wrap neonate • Breastfeed within 1 hour of delivery • Thermal care: dry and wrap neonate immediately after birth, delay first bath for 3 days • Seek trained care promptly in case of danger signs for mother or baby 0–27 days Visit by community-based worker as soon as possible • Essential newborn care: early and exclusive breastfeeding, cord care postnatal after birth to provide counselling on breastfeeding, and thermal care essential newborn care (thermal care, hygiene, clean • Apply no substance to the cord stump cord care), maternal and newborn danger signs • Detect danger signs and seek care from trained health providers. and health-care utilization; follow-up visits for sick, premature or low-birth-weight neonates.

anganwadi, government child-care and mother-care centre. data from Uttar Pradesh state only. A A baseline household survey was investigators set up an independent data quasi-experimental design was used and conducted between January and June quality assurance system that included the study design, data collection and 2003 to establish rates of programme re-interviewing 5% of households, weekly analysis were conducted by a team of coverage, maternal knowledge and comparisons of original and re-interview independent researchers who were not practices, and neonatal mortality. The data to identify disagreements, and ad- involved in the implementation of the newborn component of INHP began ditional field visits and training to data intervention. Two of the authors of in July 2003. Household surveys were collectors to resolve discrepancies. this paper (UK and DP) were involved repeated at the end of the project, in programme implementation. between January and March 2006. All Statistical analysis One INHP district, Barabanki, households in the selected sectors were Descriptive statistics were calculated served as the intervention district, while included in both surveys; respondents with use of standard methods. A house- a district receiving standard government were women who had had a live birth hold standard-of-living index was con- health and Integrated Child Develop- or stillbirth within the reference period: structed as an additive scale of house- ment Services, Unnao, was used as a calendar years 2001–2002 for baseline hold assets using the same methodology comparison district. From the 15 rural and 2004–2005 for endline. Only live as in India’s National Family Health blocks in each district, we used a com- births were included in this analysis. Survey (NFHS).20 Scores were created puter programme to randomly select Data collection included information by assigning points for household type, nine blocks in the intervention district on household and maternal charac- sanitation facilities, electricity and water and eight blocks in the comparison teristics, exposure to the intervention, supply, ownership of land and livestock, district (the difference in the number maternal and newborn care practices, and household goods. Labels of low, of blocks selected was due to differ- and pregnancy history. medium and high were assigned based ences in population size of the blocks in Data collection was contracted out on NFHS categories. χ2 analysis was Barabanki and Unnao). One sector, an to an independent survey research agency used to test for homogeneity of house- area with 15–25 anganwadi centres and (TNS India), which recruited data collec- hold socioeconomic and demographic an estimated population of 20 000 to tors with at least a 10th grade education factors and maternal characteristics be- 25 000 people, was randomly selected who were fluent in thelocal language and tween the intervention and comparison from each of the selected blocks. The dialect. The data collectors received 7 days districts at baseline and endline. sample size was calculated to detect a of didactic training and 3 days of field All indicators of exposure to the 20% reduction in neonatal mortality practice and were deployed only if they intervention and behaviour change following the intervention with 80% qualified in a post-training test. In addi- were based on survey respondents’ power at a 5% significance level. tion to the agency’s data quality assurance, self-report. Programme coverage was

798 Bulletin of the World Health Organization | October 2008, 86 (10) Research Abdullah H Baqui et al. Nutrition and health programme’s impact on neonatal mortality considered as the proportion of women estimated at 50%, we estimated that However, in the group of women who who reported being visited at home antenatal coverage was also 50%. We received a postnatal visit within 28 days, by either an auxiliary nurse-midwife, used Stata, version 8 (StataCorp. LP, the neonatal mortality rate of 35.7 deaths anganwadi worker or change agent dur- College Station, TX, United States of per 1000 live births (95% CI: 29.2–42.1) ing pregnancy (antenatal coverage) and America) for statistical analysis.24 was significantly lower than that in the during the first 28 days after delivery group who received either no visit (53.8, (postnatal coverage). Qualified doctors, Results 95% CI: 48.9–58.8) or only an ante- nurses, lady health visitors and auxiliary natal visit (54.0, 95% CI: 45.9–62.1). nurse-midwives were considered skilled Baseline sociodemographic characteris- Women who received a postnatal visit providers for antenatal, delivery and tics of study participants are shown in within 3 days of delivery had a 25% newborn care. Women were deemed Table 2. Increases in coverage were mea- lower neonatal mortality rate than did to have had a skilled birth attendant if sured in the intervention district for the those who received no visit, which was they reported giving birth at a govern- proportion of women who received only also statistically significant (Table 4). ment, NGO or private health facility an antenatal home visit (from 14% to Since neonates who died in the first few or if a skilled provider attended their 22%) and the proportion who received hours of life might have a lower prob- home delivery. Intervention exposure an antenatal and a postnatal visit within ability of receiving a postnatal visit, we and behaviour change indicators were 28 days (from 2% to 34%), while these repeated the analysis excluding deaths analysed using a difference-in-difference indicators showed little change in the that occurred on the day of birth. In this test to compare the change from base- comparison district (P < 0.001; Table 3). subset, the reduction due to postnatal line to endline for intervention versus At endline, coverage of antenatal visits visits was still statistically significant at comparison districts.21 Linear prob- was 56% for the intervention district 28%, compared to those who received ability regression models were used and 21% for the comparison district, no antenatal or postnatal visit (data with interaction terms for time (base- while coverage of postnatal visits within not shown). line versus endline) and intervention 28 days was 39% for the intervention The predicted model indicated that district to test the null hypothesis that district and 6% for the comparison increasing coverage of antenatal and the difference-in-difference equals zero, district. Those who received a home visit postnatal visits during the first 28 days controlling for maternal age, education, were slightly more likely to be Hindu, from 10% to 50%, would reduce the parity, religion and standard-of-living or to have a middle or high standard of neonatal mortality rate by 19%; increas- score. Logistic models were not used living; these differences were adjusted for ing coverage from 10% to 90% would due to the problem of using them with in subsequent analyses. yield a 34% reduction (Fig. 2; available 22 interaction terms. The behaviours that were pro- at: http://www.who.int/bulletin/vol- Neonatal deaths were those that moted through the programme showed umes/10/07-042226/en/index.html). occurred within 28 days of a live birth. greater increases in frequency in the Neonatal mortality rates – i.e. number intervention district than in the com- Discussion of neonatal deaths per 1000 live births parison district, after controlling for – were calculated separately for interven- maternal and household characteristics, We have shown that an INHP with a tion and comparison districts and 95% newborn component increased cover- although the size of the changes varied confidence intervals (CIs) were calcu- age of antenatal and postnatal health- (Table 3). For example, the proportion lated with binomial estimation method promotion visits, and improved some of households that practiced clean cord for a variance of proportion. neonatal-care practices. However, most care more than doubled in the inter- Neonatal mortality rates were also of the gain in coverage was in the ante- vention district (from 32% at baseline calculated after stratification by antena- natal period, and fewer than one in four to 68% at endline), while coverage of tal and postnatal home visitation status, neonates received a home visit within combining data from both districts. skilled birth attendance increased from the crucial first 3 days of life, when The rates were then adjusted using 16% to 22%. half of all neonatal deaths occurred.25 direct standardization 23 to account for Adjusted baseline neonatal mortality Moreover, only three of the selected differences in maternal age, education, rates did not differ between the interven- behaviours – receiving two tetanus tox- parity, religion and standard-of-living tion (46.4 deaths per 1000 live births, oid immunizations, saving money for score. Using coefficients from adjusted 95% CI: 42.0–50.8) and comparison childbirth and maintaining clean cord logistic regression, we estimated the districts (45.8 deaths per 1000 live care – were adopted by most women in marginal changes in neonatal mortality births, 95% CI: 40.6–51.0) and showed the intervention district. for various levels of coverage of ante- no evidence of change from baseline to Although the programme failed natal and postnatal home visits. The endline (Table 4). At endline, neonates to reduce neonatal mortality, those effect of antenatal visits was assessed of women in both districts who received women who received a postnatal visit by varying the antenatal coverage levels a visit from a community-based worker were significantly less likely to have from 0% to 100% assuming no post- during the antenatal period only, did not their neonate die, and our simulation natal visits had occurred. Coverage of have a lower neonatal mortality rate (54.0 analysis suggested that 90% programme postnatal visitation within 28 days was deaths per 1000 live births, 95% CI: coverage with an antenatal visit and a varied between 0% and 100% keeping 45.9–62.1) than those of women who postnatal home visit within the first antenatal coverage at the same level. received no visit (53.8 deaths per 1000 28 days could reduce the neonatal mor- For example, if postnatal coverage was live births, 95% CI: 48.9–58.8; Table 4). tality rate by 34% from baseline.

Bulletin of the World Health Organization | October 2008, 86 (10) 799 Research Nutrition and health programme’s impact on neonatal mortality Abdullah H Baqui et al.

Table 2. Selected maternal and household characteristics of recently delivered women by study district: baseline (2001–2002) and endline (2004–2005) surveys

Characteristic a Number of recently delivered women (%) Baseline survey Endline survey Comparison district Intervention district Comparison district Intervention district (n = 6 196) (n = 8 756) (n = 6 014) (n = 7 812) Maternal characteristics Age at most recent birth < 20 years 1267 (20.4) 2012 (23.0) 941 (15.7) 1366 (17.5) 20–34 years 4498 (72.6) 6000 (68.5) 4613 (76.8) 5663 (72.6) 35–49 years 431 (7.0) 742 (8.5) 453 (7.5) 775 (9.9) Education level Illiterate 3914 (63.2) 6707 (76.6) 3532 (58.7) 5845 (74.8) Literate, but middle school not completed 988 (15.9) 1161 (13.3) 915 (15.2) 991 (12.7) Middle school completed 766 (12.4) 559 (6.4) 907 (15.1) 608 (7.8) High school completed and above 528 (8.5) 329 (3.7) 660 (11.0) 368 (4.7) Parity 1 1269 (20.5) 1648 (18.8) 1313 (21.9) 1546 (19.8) 2–3 2313 (37.3) 3080 (35.2) 2401 (39.9) 2850 (36.5) 4–5 1504 (24.3) 2241 (25.6) 1396 (23.2) 2011 (25.7) ³ 6 1110 (17.9) 1787 (20.4) 904 (15.0) 1405 (18.0) Household characteristics Religion Hindu 5834 (94.2) 7128 (81.4) 5666 (94.2) 6265 (80.2) Muslim/other 362 (5.8) 1628 (18.6) 348 (5.8) 1547 (19.8) Caste Scheduled castes/tribes 2179 (43.8) 2874 (40.6) 2566 (42.7) 3166 (40.6) Other backward class b 2163 (43.5) 3371 (47.6) 2583 (43.0) 3723 (47.7) Others 634 (12.7) 834 (11.8) 858 (14.3) 1775 (11.7) Standard-of-living index score Low 3293 (53.2) 4072 (46.5) 2921 (48.6) 3122 (40.0) Medium 1977 (31.9) 3526 (40.3) 1926 (32.0) 3153 (40.3) High 926 (14.9) 1158 (13.2) 1167 (19.4) 1537 (19.7)

a c² analyses for difference in distribution between intervention and comparison at baseline and endline were all significant atP < 0.05. b As notified by central government of India who are entitled for positive discrimination and 27% of government jobs are reserved for them.

Most of the reduction in mortal- visit without a preceding antenatal visit but also develop new methods to assess ity was in the group who were visited was hampered; however, in practice the quality, coverage, and impact of within the first 3 days of birth. To antenatal visits would be necessary to programmes. check for survivor bias, we re-analysed establish contact with pregnant women This study has limitations. Al- the data excluding babies who died on to plan for postnatal visits. though CARE-India implemented the the first day of life and noted that the To our knowledge, this is the first INHP at scale in 70 districts of eight effect on mortality remained statistically study to assess the impact on mortality states of India, this evaluation was significant. The effect of postnatal visits of a large-scale community-based neo- conducted only in Uttar Pradesh. A on rates of neonatal deaths was most natal health programme implemented quasi-experimental design was selected likely due to promotion of essential with use of multipurpose government because a cluster-randomized trial was newborn care practices including early health workers. Few large-scale pro- not feasible. Assessment of behaviours and exclusive breastfeeding, thermal grammes are rigorously evaluated with was based on respondents’ report, care and clean cord care as well as mortality as the outcome. Furthermore, which might have led to some recall through identification and referral of our study was conducted by a team that error. However, an independent data neonates with signs of illness. Because was independent of programme imple- quality-assurance system was estab- the number of women who received a mentation. This study shows the value lished to improve data quality, and postnatal visit but no antenatal visit of large-scale programme evaluations the rate of recall error should not have was so small, our ability to draw con- and our findings reinforce the need differed between the intervention and clusions about the effect of a postnatal to not only validate existing methods comparison groups.

800 Bulletin of the World Health Organization | October 2008, 86 (10) Research Abdullah H Baqui et al. Nutrition and health programme’s impact on neonatal mortality

Table 3. Programme exposure and behaviour change in recently delivered women by study district: baseline (2001–2002) and endline (2004–2005)

Indicator Number of recently delivered women (%) Baseline survey Endline survey Adjusted P-value a Comparison Intervention Comparison Intervention district district district district (n = 6 196) (n = 8 756) (n = 6 014) (n = 7 812) Exposure to intervention Home visits < 0.001 b No visit 4623 (74.6) 7249 (82.8) 4673 (77.7) 2995 (38.3) Only antenatal visit 1241 (20.0) 1212 (13.9) 1010 (16.8) 1747 (22.4) Only postnatal visit within 28 days 123 (2.0) 116 (1.3) 91 (1.5) 450 (5.8) Antenatal visit and postnatal visit: within 28 days 209 (3.4) 179 (2.0) 240 (4.0) 2620 (33.5) within 3 days 99 (1.6) 79 (0.9) 119 (2.0) 1585 (20.3) < 0.001 within 7 days 129 (2.1) 111 (1.2) 149 (2.5) 2025 (25.9) < 0.001 Behaviour change Antenatal care Proportion of mothers who: Received ³ 1 antenatal check-up c 1517 (24.5) 1451 (16.6) 1653 (27.5) 2771 (35.5) < 0.001 Received ³ 3 antenatal check-ups c 426 (6.9) 316 (3.6) 538 (9.0) 952 (12.2) < 0.001 Received ³ 2 tetanus immunizations 3590 (57.9) 4189 (47.8) 3764 (62.6) 5495 (70.3) < 0.001 Consumed ³ 100 iron-folic acid tablets 412 (6.7) 436 (5.0) 497 (8.3) 1663 (21.3) < 0.001 Saved money for childbirth 758 (12.2) 1294 (14.8) 1795 (29.9) 3936 (50.4) < 0.001 Took any other birth planning step d 1232 (19.9) 933 (10.7) 1082 (18.0) 2707 (34.6) < 0.001 Delivery and newborn care Proportion of mothers who: Delivered in a health facility or at home with a skilled 1085 (17.5) 1423 (16.3) 1314 (21.8) 1756 (22.5) < 0.009 birth attendant c Practiced clean cord caree 1992 (36.0) 2471 (32.1) 2117 (41.5) 4488 (68.4) < 0.001 Practiced newborn thermal care at least for first 6 hours f 47 (0.8) 341 (3.9) 38 (0.6) 1900 (24.3) < 0.001 Initiated breastfeeding in first hour 148 (2.4) 268 (3.1) 347 (5.8) 2948 (37.7) < 0.001 Newborn check-up: c within 3 days 263 (4.2) 101 (1.2) 483 (8.0) 1297 (16.6) < 0.001 within 7 days 383 (6.2) 152 (1.7) 552 (9.2) 1529 (19.6) < 0.001 within 28 days 571 (9.2) 209 (2.4) 712 (11.8) 1887 (24.2) < 0.001

a P-value for difference-in-difference test adjusted for age, education, parity, religion and standard-of-living score. b For change in coverage of all home visits (i.e. postnatal and antenatal) from baseline to endline (comparison group versus intervention group). c From a medically qualified doctor, nurse, lady health visitor or auxiliary nurse-midwife; newborn check-up includes care sought for sick babies and well baby care. d At least one of the following: suitable location for delivery, person to deliver baby, hospital/clinic to be attended in case of complication, arrangement for transport and disposable delivery kit. e Umbilical cord cut with boiled blade and tied with sterile thread. f Neonate dried and wrapped immediately after delivery and first bath delayed for at least 6 hours.

Most deaths in children younger the intensity of services received from health services coverage.8,29–31 In previ- than 5 years – including neonatal auxiliary health workers and mortality ous evaluations of programmes from deaths – could be prevented if simple, rates.26 The multi-country evaluation the Integrated Child Development Ser- proven interventions were implemented of Integrated Management of Child- vices, nutritional supplementation to at scale.4,6 Our simulation analysis find- hood Illness (IMCI) also found varying children and mothers was increased, but ing was consistent with modelled results degrees of effectiveness,27–31 largely due only to 53% and 26%, respectively, after from the Lancet Neonatal Survival to variations in the health system con- almost 10 years of implementation.17 Series, which suggested that a package texts and how well the programme had By comparison, coverage of childhood of family-community care at 90% cov- been implemented.32 While the train- immunization reached more than 60% erage could reduce neonatal mortality ing of health workers improved quality after 5 years.34 by 18–37%.4 Seminal work on child of care,28,31,33 coverage of community We suspect that the lack of a health and nutrition in rural India also activities remained low and researchers population-level effect on mortality showed an inverse relationship between noted little effect on care-seeking or was related to the degree of scale-up

Bulletin of the World Health Organization | October 2008, 86 (10) 801 Research Nutrition and health programme’s impact on neonatal mortality Abdullah H Baqui et al. of the intervention. The intervention Table 4. Neonatal mortality rate by study district and by postnatal visit status: was scaled up to an extent in that the baseline (2001–2002) and endline (2004–2005) surveys community-based workers were present throughout the study area, and 43% Study district and Live Deaths Unadjusted Adjusted of auxiliary nurse-midwives, 91% of postnatal visit status births mortality rate per mortality rate per anganwadi workers and 98% of the 1000 live births 1000 live births a change agents received the training in (95% CI) (95% CI) neonatal care. However, scale-up was Intervention district limited in the sense that coverage was Baseline survey 8 756 431 49.2 (44.8–54.0) 46.4 (42.0–50.8) not as good as it could have been. Endline survey 7 812 393 50.3 (45.6–55.4) 52.1 (47.2–57.0) Through interviews with pro- gramme managers and community- Comparison district based workers, we identified that the Baseline survey 6 196 296 47.8 (42.6–53.4) 45.8 (40.6–51.0) workers’ competency in the new neona- Endline survey 6 014 299 49.7 (44.4–55.5) 48.6 (42.9–54.2) tal component of the programme, their Both districts workload and inadequate management Baseline survey and supervision were possible barri- No visit 11 872 588 49.5 (45.7–53.6) 47.9 (44.1–51.6) ers to higher coverage. CARE-India Antenatal visit only 2 453 123 50.1 (41.8–59.5) 44.8 (36.8–52.9) conducted an evaluation and noted Postnatal visit: that change agents’ reach was lower b than expected. Further, they identified any 627 16 25.5 (14.7–41.1) 16.0 (7.5–24.5) that anganwadi workers and auxiliary within 3 days 276 5 18.1 (5.9–41.8) 8.9 (1.2–16.7) nurse-midwives needed simple systems within 7 days 372 8 21.5 (9.3–41.9) 14.6 (4.4–24.8) to prioritize their competing job priori- Subtotal 14 952 727 48.6 (45.2–52.2) 47.1 (46.9–54.2) ties, more frequent and effective field Endline survey supervision and more intensive train- No visit 7 668 426 55.6 (50.5–60.9) 53.8 (48.9–58.8) ings with better field-based assessment Antenatal visit only 2 757 154 55.9 (47.6–65.1) 54.0 (45.9–62.1) of skills.35 Postnatal visit: Reaching newborn babies at the any b 3 401 112 32.9 (27.2–39.5) 35.7 (29.2–42.1) community level is crucial in settings within 3 days 1995 78 39.1 (31.0–48.6) 40.2 (31.8–48.6) where the availability and utilization within 7 days 2 547 92 36.1 (29.2–44.1) 39.4 (32.0–46.8) of facility-based care is low. While the Subtotal 13 826 692 50.1 (46.5–53.8) 50.6 (46.9–54.2) training of multipurpose health and nutrition workers in essential newborn a Adjusted for age, education, parity, religion and standard-of-living scores using direct standardization. b care is necessary, systems must also be Includes women who received antenatal visits and those who did not; of those receiving a postnatal visit, put in place to ensure that these work- 62% (N = 388) at baseline and 84% (N = 2860) at endline also received an antenatal visit. ers visit neonates at home during the first hours and days after birth and that the United States Agency for Interna- Research Activity Award # HRN- they can provide a link to competent tional Development (USAID) for their A-00-96-90006-00 to the Johns Hop- health services. ■ support, valuable insights and guidance. kins Bloomberg School of Public The study would not have been possible Health. Acknowledgements without the hard work and dedica- We thank the residents of the study tion of project staff of CARE-India, Competing interests: Usha Kiran and districts who gave their time generously; TNS India and the Johns Hopkins Dharmendra Panwar managed the the Government of India and CARE- University. intervention but were not directly India colleagues for their cooperation involved in the evaluation. All other and support; and special thanks to Funding: This project was funded by authors declare no conflict of interest. Massee Bateman and Neal Brandes of USAID, India Mission, through Global

Résumé Impact d’un programme intégré de nutrition et de santé sur la mortalité néonatale dans des districts ruraux du Nord de l’Inde Objectif Evaluer l’impact de la composante sanitaire d’un de l’Uttar Pradesh, au Nord de l’Inde. Les mères ayant donné programme communautaire intégré à grande échelle de nutrition naissance à un enfant dans les 2 ans précédant les enquêtes ont et de santé. été interrogées dans l’enquête de référence (n = 14 952) et dans Méthodes En faisant appel à une méthode quasi-expérimentale, l’enquête finale (n = 13 826). La principale mesure de résultat nous avons évalué un programme partenaire d’une organisation était la baisse de la mortalité néonatale. non gouvernementale et mis en place par le gouvernement Résultats Dans le district d’intervention, la fréquence des visites à indien dans les infrastructures existantes de deux districts ruraux domicile par des agents de santé communautaires a augmenté à

802 Bulletin of the World Health Organization | October 2008, 86 (10) Research Abdullah H Baqui et al. Nutrition and health programme’s impact on neonatal mortality la fois pendant la période anténatale (de 16 à 56 %) et postnatale visite postnatale (53,8 décès pour 1000 naissances vivantes, IC (de 3 à 39 %), tout comme la fréquence des soins apportés à la à 95 % : 48,9-58,8), après ajustement en fonction des variables mère et au nouveau-né. Dans le district de comparaison, aucune sociodémographiques. Les trois-quarts de la réduction de mortalité progression des visites à domicile n’a été observée et le seul ont été relevés chez les nouveau-nés visités dans les 3 jours suivant changement comportemental notable a été la mise de côté par la naissance. Cet effet sur la mortalité restait statistiquement les femmes d’argent destiné aux traitements médicaux d’urgence. significatif lorsqu’on excluait les décès de nourrissons intervenus Les taux de mortalité néonatale restaient inchangés dans les deux le jour de la naissance. districts tant que les nouveau-nés n’avaient reçu qu’une visite Conclusion La couverture limitée du programme ne permettait anténatale. Toutefois, les nouveau-nés ayant reçu une visite pas d’observer un effet sur la mortalité néonatale à l’échelle postnatale à domicile dans les 28 jours suivant la naissance d’une population. La réduction constatée du taux de mortalité des présentaient un taux de mortalité néonatale inférieur de 34 % (35,7 nouveau-nés ayant reçu des visites domiciliaires postnatales décès pour 1000 naissances vivantes, intervalle de confiance à prouve la capacité du programme à influer sur la mortalité 95 % : 29,2-42,1) à celui des nouveaux-nés n’ayant pas reçu de néonatale.

Resumen Impacto de un programa integrado de nutrición y salud en la mortalidad neonatal en una zona rural del norte de la India Objetivo Evaluar los efectos del componente de salud neonatal de de mortalidad neonatal se mantuvieron inalteradas en ambos un programa comunitario en gran escala que integraba medidas distritos en los casos en que sólo se recibió una visita de atención de nutrición y de salud. prenatal. Sin embargo, entre los recién nacidos que recibieron una Métodos Utilizando un diseño cuasiexperimental, evaluamos los visita domiciliaria de atención posnatal en sus primeros 28 días resultados de un programa promovido por una organización no de vida se observó una disminución del 34% de la mortalidad gubernamental y aplicado por el Gobierno de la India en el marco neonatal (35,7 defunciones por 1000 nacidos vivos, intervalo de de la infraestructura existente en dos distritos rurales de Uttar confianza (IC) del 95%: 29,2–42,1) en comparación con los que Pradesh, en el norte del país. Se entrevistó a madres que habían no recibieron ninguna visita posnatal (53,8 defunciones por 1000 dado a luz en los dos años que precedieron a las encuestas (n = nacidos vivos, IC95%: 48,9–58,8), tras ajustar en función de 14 952 en la encuesta basal, y n = 13 826 en la encuesta final). las variables sociodemográficas. Las tres cuartas partes de esa La variable de evaluación principal fue la reducción de la mortalidad reducción de la mortalidad se explican por los niños que recibieron neonatal. la visita durante los tres primeros días posteriores al nacimiento. El Resultados En el distrito de intervención, las visitas domiciliarias efecto en la mortalidad seguía siendo estadísticamente significativo por agentes comunitarios aumentaron tanto durante el periodo tras excluir del análisis a los niños fallecidos el mismo día en que prenatal (del 16% al 56%) como durante el periodo posnatal (del nacieron. 3% al 39%), al igual que la frecuencia de prácticas de atención de Conclusión La limitada cobertura del programa impide determinar la madre y el recién nacido. En el distrito de comparación no se cualquier efecto en la mortalidad neonatal a nivel poblacional. observó mejora alguna de las visitas domiciliarias, y el único cambio La reducción de las tasas de mortalidad neonatal entre quienes de comportamiento destacable fue que las mujeres habían ahorrado recibieron visitas domiciliarias posnatales demuestra el potencial dinero para un posible tratamiento médico de urgencia. Las tasas del programa para influir en las defunciones neonatales.

ملخص تأثري التغذية املتكاملة والربنامج الصحي عىل وفيات الولدان يف املناطق الريفية الواقعة شامل الهند الغرض:تقيـيم تأثري مكون صحة الولدان يف برنامج مجتمعي واسع النطاق السيدات لألموال ُّتحسباً للمعالجة الطبية الطارئة. وظلت ّمعدالت وفيات للتغذية املتكاملة والصحة. الولدان يف َاملنطقت نْي دون تغيـري، بيد أن الولدان الذين تلقوا زيارات منزلية الطريقة: تم استخدام تصميم تجربة غري كاملة يف تقيـيم برنامج نفذته تالية للوالدة خالل 28 يوما ًمن امليالد، شهدوا انخفاضا ًيف معدل وفيات الحكومة الهندية مبساعدة املنظامت الالحكومية، يف َمنطقت نْي َريفيت نْي يف الولدان بنحو 34% )35.7 وفاة لكل 1000 وليد حي، بفاصل ثقة 95%، إذ أوتار برادش، شامل الهند، وذلك باالعتامد عىل البنية األساسية الحالية. تراوح ّمعدل األرجحية بي 29.2 – 42.1(، باملقارنة مع من مل يتلقوا زيارات جريتُوأ لقاءات مع األمهات الاليت وضعن حملهن يف َالعام نْي َالسابق نْي إلجراء تالية للوالدة )53.8 وفاة لكل 1000 وليد حي، بفاصل ثقة 95%، إذ تراوح املسوحات، أثناء املسوحات يف بداية األنشطة )العدد = 14952( واملسوحات ّمعدل األرجحية بي 48.9 و58.8(،مبا يتوافق مع املتغريات االجتامعية يف نهاية األنشطة )العدد = 13826(. وكان قياس النتيجة األولية هو تخفيض والدميوغرافية. وقد وقعت ثالثة أرباع االنخفاض يف الوفيات بي األطفال وفيات الولدان. الذينمتت زيارتهم يف األيام الثالثة األوىل من امليالد. وقد ظل األثر عىل النتائج: شهدت منطقة املداخالت زيادة الزيارات املنزلية من ِق َبل العاملي الوفيات يعتد به إحصائيا ًعند استبعاد األطفال الذين ماتوا يف يوم الوالدة. ِّاملجتمعيـي يف كل من الفرتة السابقة للوالدة )من 16% إىل 56%( ويف االستنتاج:إن التغطية املحدودة للربنامج مل متكن من مالحظة األثر عىل الفرتة التالية للوالدة )من 3% إىل 39%(، وكذلك معدل تكرار مامرسات وفيات الولدان عىل مستوى السكان. إن انخفاض معدالت وفيات الولدان رعاية األمهات والولدان. أما املنطقة األخرى محل املقارنة، فلم نلحظ أي بي من ّيتلقون زيارات منزلية تالية للوالدة، يظهر إمكانية تأثري الربنامج عىل ّتحسنيف الزيارات املنزلية، وأن التغري السلويك الوحيد امللحوظ هو ادخار وفيات الولدان.

Bulletin of the World Health Organization | October 2008, 86 (10) 803 Research Nutrition and health programme’s impact on neonatal mortality Abdullah H Baqui et al.

References 1. Lawn JE, Cousens S, Zupan J. 4 million neonatal deaths: When? Where? 20. National Family Health Survey-II (NFHS-II) 1989-1999. Mumbai: International Why? Lancet 2005;365:891-900. PMID:15752534 doi:10.1016/ Institute for Populations Sciences and ORC-Macro; 2001. S0140-6736(05)71048-5 21. Ganzach Y, Saporta I, Weber Y. Interaction in linear versus logistic models: 2. Improving neonatal health in south-east asia region: report of a regional a substantive illustration using the relationship between motivation, consultation. New Delhi: WHO; 2002. ability, and performance. Organ Res Methods 2000;3:237-52. 3. Black RE, Morris SS, Bryce J. Where and why are 10 million children dying doi:10.1177/109442810033002 every year? Lancet 2003;361:2226-34. PMID:12842379 doi:10.1016/ 22. Ai C, Norton EC. Interaction terms in logit and probit models. Econ Lett S0140-6736(03)13779-8 2003;80:123-9. doi:10.1016/S0165-1765(03)00032-6 4. Darmstadt GL, Bhutta ZA, Cousens S, Adam T, Walker N, de Bernis L. 23. Rothman K. Epidemiology: an introduction. New York: Oxford University Press; Evidence-based, cost-effective interventions: how many newborn babies can 2002. we save? Lancet 2005;365:977-88. PMID:15767001 doi:10.1016/S0140- 24. Stata Statistical Software: Release 8. College Station, TX: StataCorp. LP; 6736(05)71088-6 2003. 5. Darmstadt GL, Lawn JE, Costello A. Advancing the state of the world’s 25. Baqui AH, Darmstadt GL, Williams E, Kumar V, Kiran T, Panwar D, et al. newborns. Bull World Health Organ 2003;81:224-5. PMID:12764520 Rates, timing and causes of neonatal deaths in rural India: implications for 6. Jones G, Steketee RW, Black RE, Bhutta ZA, Morris SS. How many child neonatal health programmes. Bull World Health Organ 2006;84:706-13. deaths can we prevent this year? Lancet 2003;362:65-71. PMID:12853204 PMID:17128340 doi:10.2471/BLT.05.026443 doi:10.1016/S0140-6736(03)13811-1 26. Kielmann AA, Taylor CE, Parker RL. The Narangwal Nutrition Study: a 7. Bhutta ZA, Darmstadt GL, Hasan BS, Haws RA. Community-based summary review. Am J Clin Nutr 1978;31:2040-57. PMID:102180 interventions for improving perinatal and neonatal health outcomes in 27. Ali M, Asefaw T, Byass P, Beyene H, Pedersen FK. Helping northern Ethiopian developing countries: a review of the evidence. Pediatrics 2005;115 communities reduce childhood mortality: population-based intervention trial. Suppl;519-617. PMID:15866863 Bull World Health Organ 2005;83:27-33. PMID:15682246 8. Bryce J, Arifeen SE, Pariyo G, Lanata C, Gwatkin D, Habicht JP. Reducing 28. El Arifeen S, Blum LS, Hoque DM, Chowdhury EK, Khan R, Black RE, et al. child mortality: can public health deliver? Lancet 2003;362:159-64. Integrated Management of Childhood Illness (IMCI) in Bangladesh: early PMID:12867119 doi:10.1016/S0140-6736(03)13870-6 findings from a cluster-randomised study.Lancet 2004;364:1595-602. 9. Bang AT, Bang RA, Reddy HM. Home-based neonatal care: summary and PMID:15519629 doi:10.1016/S0140-6736(04)17312-1 applications of the field trial in rural Gadchiroli, India (1993 to 2003). 29. Huicho L, Davila M, Campos M, Drasbek C, Bryce J, Victora CG. Scaling J Perinatol 2005;25 Suppl 1;S108-22. PMID:15791272 doi:10.1038/ up integrated management of childhood illness to the national level: sj.jp.7211278 achievements and challenges in Peru. Health Policy Plan 2005;20:14-24. 10. Bartlett AV, Paz de Bocaletti ME, Bocaletti MA. Neonatal and early PMID:15689426 doi:10.1093/heapol/czi002 postneonatal morbidity and mortality in a rural Guatemalan community: the 30. Huicho L, Davila M, Gonzales F, Drasbek C, Bryce J, Victora CG. importance of infectious diseases and their management. Pediatr Infect Dis J Implementation of the Integrated Management of Childhood Illness strategy in 1991;10:752-7. PMID:1945577 Peru and its association with health indicators: an ecological analysis. Health 11. Berggren GG, Berggren W, Verly A, Garnier N, Peterson W, Ewbank D, et al. Policy Plan 2005;20 Suppl 1;i32-41. PMID:16306067 doi:10.1093/heapol/ Traditional midwives, tetanus immunization, and infant mortality in rural Haiti. czi052 Trop Doct 1983;13:79-87. PMID:6679403 31. Bryce J, Victora CG, Habicht JP, Black RE, Scherpbier RW. Programmatic 12. Daga SR, Daga AS, Dighole RV, Patil RP, Dhinde HL. Rural neonatal care: pathways to child survival: results of a multi-country evaluation of Integrated Dahanu experience. Indian Pediatr 1992;29:189-93. PMID:1592499 Management of Childhood Illness. Health Policy Plan 2005;20 Suppl 1;i5-17. 13. Manandhar DS, Osrin D, Shrestha BP, Mesko N, Morrison J, Tumbahangphe PMID:16306070 doi:10.1093/heapol/czi055 KM, et al. Effect of a participatory intervention with women’s groups 32. Victora CG, Schellenberg JA, Huicho L, Amaral J, El Arifeen S, Pariyo G, et al. on birth outcomes in Nepal: cluster-randomised controlled trial. Lancet Context matters: interpreting impact findings in child survival evaluations. 2004;364:970-9. PMID:15364188 doi:10.1016/S0140-6736(04)17021-9 Health Policy Plan 2005;20 Suppl 1;i18-31. PMID:16306066 doi:10.1093/ 14. Pratinidhi A, Shah U, Shrotri A, Bodhani N. Risk approach strategy in neonatal heapol/czi050 care. Bull World Health Organ 1986;64:291-7. PMID:3488845 33. Armstrong Schellenberg J, Bryce J, de Savigny D, Lambrechts T, Mbuya C, 15. Jokhio AH, Winter HR, Cheng KK. An intervention involving traditional birth Mgalula L, et al. The effect of Integrated Management of Childhood Illness on attendants and perinatal and maternal mortality in Pakistan. N Engl J Med observed quality of care of under-fives in rural Tanzania.Health Policy Plan 2005;352:2091-9. PMID:15901862 doi:10.1056/NEJMsa042830 2004;19:1-10. PMID:14679280 doi:10.1093/heapol/czh001 16. Darmstadt G, Kumar V, Mohanty S, Kumar A, Mishra R, Santosham M, et al. 34. Tandon BN, Gandhi N. Immunization coverage in India for areas served by Impact of community-based behaviour change management on neonatal the Integrated Child Development Services programme. The Integrated Child mortality: a cluster-randomized, controlled trial in Shivgarh, Uttar Pradesh, Development Services Consultants. Bull World Health Organ 1992;70:461-5. India. Lancet 2008. In press. PMID:1394779 17. Avsm YS, Gandhi N, Tandon BN, Krishnamurthy KS. Integrated Child 35. Rachna program 2001-2006: summary of approaches and results. New Development Services Scheme and nutritional status of Indian children. J Delhi: Care; 2007. pp. 1-46. Trop Pediatr 1995;41:123-8. PMID:7776397 18. Nanda P. Female health workers – responsibilities and constraints. Health Millions 1993;1:25-6. PMID:12286354 19. Tandon BN, Ramachandran K, Bhatnagar S. Integrated child development service in India: evaluation of the delivery of nutrition and health services and the effect on the nutritional status of the children. Indian J Med Res 1981;73:385-94. PMID:7275234

804 Bulletin of the World Health Organization | October 2008, 86 (10) Research Abdullah H Baqui et al. Nutrition and health programme’s impact on neonatal mortality

Fig. 2. Predicted neonatal mortality rates varying antenatal and postnatal programme coverage equally a

60 55.6 52.7 50.1 50 47.5 45.1 42.8 40.6 40 38.5 36.5 34.6 32.8 30

20 (per 1000 live births)

Neonatal mortality rate 10

0 0 10 20 30 40 50 60 70 80 90 100 Antenatal and postnatal programme coverage (%)

a Using coefficients from adjusted logistic regression, the marginal changes in neonatal mortality were estimated for various levels of coverage of antenatal and postnatal home visits; the effect of antenatal visits was assessed by varying the antenatal coverage levels from 0% to 100% assuming no postnatal visitation. Coverage of postnatal visitation within 28 days was varied between 0% and 100% keeping antenatal coverage at the same level. For example, if postnatal coverage was estimated at 50%, antenatal was also estimated at 50%.

Bulletin of the World Health Organization | October 2008, 86 (10) A